Skip to content

Fun Fact: The Voice Stephen Hawking Refused to Upgrade

Abstract

Stephen Hawking spoke through a 1980s American speech synthesizer for the last three decades of his life, and turned down every offer of a cleaner, more modern, more British-sounding voice. The accent was an accident of the hardware he happened to be handed in the 1980s. Keeping it was a choice. “I keep it because I have not heard a voice I like better,” he explained, “and because I have identified with it.” By the end the electronics were so old that a team had to reverse-engineer the circuit and emulate it on a Raspberry Pi so the world would not lose the voice.

Where the Voice Came From

The sound traces back to Dennis Klatt, an MIT researcher who spent from the mid-1960s onward modeling how the human vocal tract produces speech. His formant-synthesis system, known both as KlattTalk and MITalk, generated speech from rules about sound production rather than by splicing recordings. Digital Equipment Corporation licensed it and shipped it in 1984 as DECtalk, a synthesizer with a roster of named voices (nine by its later versions). The default was named “Perfect Paul,” modeled partly on Klatt’s own voice.

Hawking’s machine was not DECtalk itself but a close relative: the CallText 5010 from a company called Speech Plus, built on the same Klatt lineage and used by Hawking from 1988. It produced a crisp, American-accented voice. Hawking lost his own speech in 1985 after an emergency tracheotomy following pneumonia. He first spelled words letter by letter, then received the Equalizer software written by Walter Woltosz, who had built it for his mother-in-law with ALS. The CallText synthesizer was wired into Hawking’s wheelchair, and the combination became the voice the public knew.

Keeping It on Purpose

As synthesis improved, Hawking was repeatedly offered smoother voices, including ones with a British accent to match the man. He refused all of them. The 1980s synthesizer robbed his speech of intonation and gave an English physicist an American twang, and he kept it precisely because it had become him. The voice was distinctive enough that Hawking treated it as part of his identity, and reportedly resisted anyone “correcting” the accent to sound more like where he was from.

The loyalty created an engineering problem. The CallText 5010 hardware aged toward failure and spare parts vanished. Rather than swap in a modern voice, Hawking’s team and outside engineers reverse-engineered the old synthesizer and reproduced its exact sound in software, running on a Raspberry Pi that he used from January 2018 until his death that March. The point of the exercise was not to make the voice better. It was to make sure it stayed exactly the same.

The story sits at the intersection of assistive technology and identity: a piece of assistive computing so bound up with the person using it that its owner defended a decades-old accent as his own. See also the broader history of machine speech in The Voice Assistant Revolution.

📚 Sources